High Kurtosis/Skewness -> problem?

[Zytozid]

Question: One of my constructs/ explaining variables is not significant for any of its path coefficients and I wonder whether this result could be becaue of the data being skewed/having a high kurtosis.

Any idea whether this is problematic or should I "trust" the result of my Bootstrap? Generally: What implication does a high kurtosis/skewness of my variables for my results and their interpretation?


The info:

One of my independent constructs ("fct") is measured by 3 reflective items. All items have a relatively high a) (negative!) skewness and b) (positive) kurtosis.

In my model "fct" impacts two variables, lets call them"att" and "beh".

"att" also impacts "beh" and finally, "beh" is the only explaining variable for "use" which is the final DV.

There are 5 other explaining constructs next to "fct".

Infos to the items of "fct":

fct1: a) -1,39 b) 3,01

fct2: a) -1,18 b) 1,95

fct3: a) -1,05 b) 1,87

N = 397, I have many respondents. Using Smart PLS and SPSS.


Skewness/ Kurtosis for "att" and "beh":

att1 -1,1 / 1,4

att2 -1,5 / 2,5

att3 -1,2 / 1,4

att4 - 1,4 / 1,8



beh1 -2,1 / 5,7 (!)

beh2 -1,3 / 1

[cringle]

A non-siginficant results can certainly occur. But in any case, you should make sure that your sample is not too small and your data is not problematic in any other way (this is independent from any analysis method).

In SmartPLS, we suggest running the bias-corrected and accelerated bootstrap procedure. Then, check the bias-corrected bootstrap confidence intervals in the results report.

Best regards
Christian